Generative Adversarial Nets from a Density Ratio Estimation Perspective

Abstract

Generative adversarial networks (GANs) are successful deep generative models. GANs are based on a two-player minimax game. However, the objective function derived in the original motivation is changed to obtain stronger gradients when learning the generator. We propose a novel algorithm that repeats the density ratio estimation and f-divergence minimization. Our algorithm offers a new perspective toward the understanding of GANs and is able to make use of multiple viewpoints obtained in the research of density ratio estimation, e.g. what divergence is stable and relative density ratio is useful.

Add contents especially theoretical things for ICLR 2017

References in corpus (8)

• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
• Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks
• Estimating divergence functionals and the likelihood ratio by convex risk minimization
• How (not) to Train your Generative Model: Scheduled Sampling, Likelihood, Adversary?
• Composite Binary Losses
• Theory and Applications of Proper Scoring Rules
• On distinguishability criteria for estimating generative models
• Minimum scoring rule inference

Cited by in corpus (29)

• Generative Adversarial Networks: An Overview
• Spectral Normalization for Generative Adversarial Networks
• Conditional Image Synthesis With Auxiliary Classifier GANs
• Variational Approaches for Auto-Encoding Generative Adversarial Networks
• Learning in Implicit Generative Models
• Variational Inference using Implicit Distributions
• Gradient Estimators for Implicit Models
• Improved generator objectives for GANs
• Likelihood-free MCMC with Amortized Approximate Ratio Estimators
• Global Convergence to the Equilibrium of GANs using Variational Inequalities
• Subsampling Generative Adversarial Networks: Density Ratio Estimation in Feature Space with Softplus Loss
• Adversarial Message Passing For Graphical Models
• Characterizing and Avoiding Negative Transfer
• Off-Dynamics Reinforcement Learning: Training for Transfer with Domain Classifiers
• C-Learning: Learning to Achieve Goals via Recursive Classification
• Discriminator Contrastive Divergence: Semi-Amortized Generative Modeling by Exploring Energy of the Discriminator
• A Theory of Label Propagation for Subpopulation Shift
• Understanding the Effectiveness of Lipschitz-Continuity in Generative Adversarial Nets
• Expected Information Maximization: Using the I-Projection for Mixture Density Estimation
• Wasserstein-Wasserstein Auto-Encoders
• Cycle-Consistent Adversarial Learning as Approximate Bayesian Inference
• Meta-Learning for Relative Density-Ratio Estimation
• Latent Dirichlet Allocation in Generative Adversarial Networks
• Importance Weighted Hierarchical Variational Inference
• 3D Human motion anticipation and classification
• BreGMN: scaled-Bregman Generative Modeling Networks
• Robust GANs against Dishonest Adversaries
• Bridging the Gap Between $f$-GANs and Wasserstein GANs
• Continual Density Ratio Estimation in an Online Setting